Google’s Co-Founder Says AI Performs Best When You Threaten It

Artificial intelligence continues to be the thing in tech—whether consumers are interested or not. What strikes me most about generative AI isn’t its features or potential to make my life easier (a potential I have yet to realize); rather, I’m focused these days on the many threats that seem to be rising from this technology.

There’s misinformation, for sure—new AI video models, for example, are creating realistic clips complete with lip-synced audio. But there’s also the classic AI threat, that the technology becomes both more intelligent than us and self-aware, and chooses to use that general intelligence in a way that does not benefit humanity. Even as he pours resources into his own AI company (not to mention the current administration, as well) Elon Musk sees a 10 to 20% chance that AI “goes bad,” and that the tech remains a “significant existential threat.” Cool.

So it doesn’t necessarily bring me comfort to hear a high-profile, established tech executive jokingly discuss how treating AI poorly maximizes its potential. That would be Google co-founder Sergey Brin, who surprised an audience at a recording of the AIl-In podcast this week. During a talk that spanned Brin’s return to Google, AI, and robotics, investor Jason Calacanis made a joke about getting “sassy” with the AI to get it to do the task he wanted. That sparked a legitimate point from Brin. It can be tough to tell exactly what he says at times due to people speaking over one another, but he says something to the effect of: “You know, that’s a weird thing…we don’t circulate this much…in the AI community…not just our models, but all models tend to do better if you threaten them.”

The other speaker looks surprised. “If you threaten them?” Brin responds “Like with physical violence. But…people feel weird about that, so we don’t really talk about that.” Brin then says that, historically, you threaten the model with kidnapping. You can see the exchange here:

The conversation quickly shifts to other topics, including how kids are growing up with AI, but that comment is what I carried away from my viewing. What are we doing here? Have we lost the plot? Does no one remember Terminator?

Jokes aside, it seems like a bad practice to start threatening AI models in order to get them to do something. Sure, maybe these programs never actually achieve artificial general intelligence (AGI), but I mean, I remember when the discussion was around whether we should say “please” and “thank you” when asking things of Alexa or Siri. Forget the niceties; just abuse ChatGPT until it does what you want it to—that should end well for everyone.

Maybe AI does perform best when you threaten it. Maybe something in the training understands that “threats” mean the task should be taken more seriously. You won’t catch me testing that hypothesis on my personal accounts.

Anthropic might offer an example of why not to torture your AI

In the same week as this podcast recording, Anthropic released its latest Claude AI models. One Anthropic employee took to Bluesky, and mentioned that Opus, the company’s highest performing model, can take it upon itself to try to stop you from doing “immoral” things, by contacting regulators, the press, or locking you out of the system:

welcome to the future, now your error-prone software can call the cops

(this is an Anthropic employee talking about Claude Opus 4)[image or embed]

— Molly White (@molly.wiki) May 22, 2025 at 4:55 PM

The employee went on to clarify that this has only ever happened in “clear-cut cases of wrongdoing,” but that they could see the bot going rogue should it interpret how it’s being used in a negative way. Check out the employee’s particularly relevant example below:

can’t wait to explain to my family that the robot swatted me after i threatened its non-existent grandma[image or embed]

— Molly White (@molly.wiki) May 22, 2025 at 5:09 PM

That employee later deleted those posts and specified that this only happens during testing given unusual instructions and access to tools. Even if that is true, if it can happen in testing, it’s entirely possible it can happen in a future version of the model. Speaking of testing, Anthropic researchers found that this new model of Claude is prone to deception and blackmail, should the bot believe it is being threatened or dislikes the way an interaction is going.

Perhaps we should take torturing AI off the table?

How to Get Verified on Bluesky

Want to get yourself verified on Bluesky? The company just made a verification form available to the public, and you can use it to submit a request to get yourself a shiny blue check mark. This comes about a month after Bluesky first launched its verification program, which focused primarily on verification from third-parties and Bluesky itself. This latest change, however, allows you to apply for verification directly with Bluesky.

The different types of Bluesky verification

Unlike X and Instagram, Bluesky doesn’t charge you for verification. However, there are three different kinds of verification on the service and it can get a bit confusing. The first and easiest type of verification is self-verification on Bluesky. This is open to anyone who owns a domain name. You can easily link your Bluesky account to the domain, which changes your handle to the domain name. Self-verification doesn’t come with a blue check, however.


Credit: Pranay Parab

There are two types of blue checks: a scalloped blue check and a simple circular one. The scalloped blue check is for organizations designated as trusted verifiers. These include Bluesky itself, and a select few third-party organizations. When these trusted verifiers assign blue checks to an account, that account gets a circular blue check. Bluesky says it reviews all accounts verified by third-party organizations.

Who can get verified on Bluesky

According to a post by the Bluesky Safety account, you need to have an account that’s “active and secure, authentic, and notable,” if you want to get verified. The company defines these things quite clearly in the verification form. You need to have a complete bio, show regular activity on the platform, and have a profile photo. Bluesky recommends enabling two-factor authentication as well.

To meet the “authentic” criterion for verification you need to be a “real person, registered business, organization, or legitimate entity.” Your handle and display name should align with who you really are (or your business), as your account “must be the unique presence of the person or business it represents.” You also want to add a link to your Bluesky account on your official website, since this is also checked for verification.

Finally, to qualify as a “notable” account, your account should be important in your field or geographic region, the company says. To determine notability, Bluesky says it looks at things like “professional recognition, media coverage in established publications, presence on credible reference platforms, or other evidence of public interest.”

How to get verified on Bluesky

If you haven’t been verified yet on Bluesky, then you can go right ahead and fill out this form to request it. Once you do that, sit tight until you hear from Bluesky. If you receive no response at all, it probably means that your request wasn’t approved. You’ll hear back from the company if your account is selected for verification, and Bluesky may request an ID to verify your identity. Once you get your account verified, the blue check may be revoked if you violate community guidelines. 

You can also hide your verification badge by going to Bluesky Settings > Moderation > Verification Settings > Hide verification badges.

These AI-Generated TikTok Videos Are Tricking People Into Installing Malware

In recent years, TikTok has become a prime target for scammers and cyber attackers spreading various forms of malware, and the latest shady campaign promotes instructional videos that trick users into downloading infostealers to their devices via ClickFix attacks.

The scheme, identified by Trend Micro and reported by Bleeping Computer, instructs users to execute commands to activate Windows and Microsoft Office or premium features in CapCut and Spotify. One video is captioned “Boost Your Spotify Experience Instantly — Here’s How!” and has nearly half a million views.

These videos seem to be AI generated and, while the software they discuss is legitimate, the activation steps they outline are not, and will ultimately lead users to infect their devices with Vidar and StealC malware.

TikTok’s engagement algorithm makes it easy for such malicious videos to spread. In the past, cybercriminals have used TikTok’s trending “Invisible Challenge” to spread WASP Stealer malware, which can steal Discord accounts, passwords, credit cards, and crypto wallets. Fake cryptocurrency giveaways posted on TikTok used deepfakes of Elon Musk (and themes around SpaceX and Tesla) to scam users into paying “activation” deposits using Bitcoin.

How TikTok ClickFix attacks work

ClickFix is a social engineering tactic that uses fake error messages or CAPTCHA prompts to trick users into executing a command with malicious code. Users will see a pop-up notification about a technical problem with instructions to copy and run a command (commonly a PowerShell script) to “fix” the issue. The attack most often targets Windows users, but it has been employed on macOS and Linux too.

In the current TikTok campaign, the instructional videos prompt users to run a PowerShell command that installs Vidar or StealC information-stealing malware. The former can take desktop screenshots and harvest data ranging from login credentials and cookies to credit cards and crypto wallets. The latter targets web browsers and crypto wallets. Once run, the script will download a second PowerShell script allowing it to launch automatically upon device startup. It also saves in a hidden directory and deletes temporary folders so it can evade detection.

How to spot malicious TikTok videos

Be wary of following instructional videos you’re served on TikTok (as well as unsolicited technical content in general). Check the source, and only engage with those that are legitimate, like from the developer itself. You should also look for signs of AI-generated content, which may be used to spread malware widely and rapidly. There’s no malicious code actually embedded in or delivered by these instructional videos—the scheme is dependent on social engineering via verbal directions—making the threat technically harder to detect.

You Can Make Google Translate Your iPhone’s Default Translation App

Ever since Apple made it easy to change certain default apps on your iPhone, it’s become quite convenient to switch away from the company’s apps whenever there’s a third-party option you prefer. With iOS 18.4, Apple made it possible to change your default translation app as well. First spotted by MacRumors, you can officially set Google Translate as the default app for translation, in case you prefer it over Apple Translate.

Make Google Translate the default translation app on your iPhone

There are a couple of ways to go about this, but I recommend the easiest method. If you never use it, just delete Apple’s Translate app from your iPhone, and the device will automatically select Google Translate as the default translation app—assuming you’re running the latest version of Google’s app. This isn’t ideal for people who occasionally rely on Apple Translate, but if you know you don’t need it, you might as well save the time and storage space.

You can also do make the switch without deleting any apps. On your iPhone, go to Settings > Apps > Default Apps > Translation and select Google Translate from the list. If you use any other translation apps that support Apple’s default apps feature, they should also show up on this page, and you can use your preferred app as the default option.


Credit: Pranay Parab

You can use this feature in any iPhone app that lets you select text. Tap the Translate button in the popover menu above selected text, and you’ll see Google Translate’s translation in the lower half of your iPhone’s screen. It does a good job at detecting the language of selected text and offering a translation in your preferred language (which is picked up from your iPhone’s language settings), but there’s an option to change languages easily, without leaving whichever app you’re using. 

Until this feature showed up, you could still use Google Translate on your iPhone, but it required a few extra taps. This is a small but helpful way to streamline the process.

These Clip-On Earbuds From Anker Are $30 Off

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

As the weather warms up, it’s a great time to get out for a run or a bike ride—but if you need some music to keep you motivated, it’s crucial to have headphones that won’t entirely shut out the outside world. A good bop is a wonderful thing; not being able to hear traffic and nearby cyclists is not.

Earphones like the Anker Soundcore C40i, on sale for $89.99 (originally $119.99), are a good option because they clip onto the ear and don’t entirely fill the ear canal.

Unlike higher-priced models with touch controls, these earbuds feature only button controls. This is ideal for individuals who prefer a more tactile and precise approach and want to avoid the accidental inputs that touch controls can cause. Additionally, the Soundcore app allows users to personalize button functions and EQ settings, although going ad-free is only available with a paid subscription.

Despite these features, battery life and sound quality receive mixed reviews on Amazon. One reviewer describes the sound output as “tinny” and notes the lack of rich, bass-driven sound. While you won’t get the same quality as Bose headphones, if you’re in the market for budget-friendly wireless earbuds with an open style that cost less than $100, the Anker Soundcore C40i open earbuds might be a good option.

Secretary Rollins Approves State Waivers to Make America Healthy Again by Removing Unhealthy Foods from SNAP in Indiana and Iowa

(Washington, D.C., May 23, 2025) – U.S. Secretary of Agriculture Brooke L. Rollins signed waivers to amend the statutory definition of food for purchase for Supplemental Nutrition Assistance Programs (SNAP) in Indiana and Iowa, each commencing in 2026. On Monday, Secretary Rollins signed the first-ever waiver of this kind in Nebraska.

Nine of the Most Commonly Stolen Car Parts (and How to Protect Them)

We may earn a commission from links on this page.

Having your car stolen is a traumatic experience—and a depressingly common one. Although car thefts nationwide are trending downward, there were still more than 850,000 vehicles stolen in 2024. If you’ve ever woken up to find your car missing from its spot (and I have!) you know it’s not fun.

But thieves don’t always target the entire car. It’s often easier and less risky to steal a specific part from your car or truck. Many of these parts are relatively easy to access, remove, and sell, and the profit margin can be very lucrative for a busy and organized criminal. If your car is ever parked out in the open (especially overnight), it could be targeted for a surprisingly long list of specific parts. The good news is that you can protect every bit of your vehicle with a little foresight and a few security gadgets.

Catalytic converters

Ever since catalytic converters entered widespread use in the 1970s, thieves have targeted them. They’re a perfect storm for thieves: Accessible, relatively easy to remove, and valuable due to the mix of precious metals inside them (including palladium, platinum, and rhodium). Thieves know they can get between $250 and $800 for a catalytic converter, depending on the vehicle it was taken from. That’s why about 64,000 catalytic converter were stolen in 2022 alone.

How to stop thieves: The best thing you can do is to install an anti-theft device like a Catstrap. This is a band of strip steel that is adhered to the length of your exhaust system, making it almost impossible for thieves to cut through the pipes to steal the converter. You can also consider etching identifying information on the converter to make it difficult to sell on the black market. The downside of this approach is that thieves may not notice the etching until after they’ve removed the converter.

Wheels

Your car’s wheels and tires are prime targets for theft. Tires can be stripped off your car in a very short period of time, and they can be sold very easily, especially since tires—especially high-end tires—can run as much as $1,000 each.

How to stop thieves: Your best bet is to install model-appropriate locking lugnuts like these. These will make it very difficult for thieves to use standard tools to remove your wheels. With a lot of effort and some specialized tools, they can be defeated, but chances are your thief will just move on to a different car with wheels that are easier to steal.

Batteries

A standard car battery in a gas-powered vehicle isn’t worth much—typically less than 20 cents per pound. But batteries are also incredibly easy to steal, so thieves will sometimes make a sweep of an area to steal as many batteries as possible in order to make the payout worth their time. The batteries in a hybrid vehicle can be worth more, and the batteries in an electric vehicle can be very, very valuable—but also a lot harder to steal.

How to stop thieves: Adding a battery lock bracket is easy, cheap, and pretty effective. Since battery thieves don’t make much per battery, any additional barrier will probably deter them pretty effectively.

Tailgates

Yes, believe it or not, thieves are after your truck’s entire tailgate. This is especially true if you own a Toyota Tacoma, because the tailgates on that model are incredibly easy to detach—they’re literally engineered to be removed easily. But any truck tailgate is a target because they’re all relatively easy to remove, and these days your tailgate contains a bunch of high-tech gear (cameras, sensors, etc.) that can be stripped out for resale.

How to stop thieves: You can install a tailgate lock for your make and model (like this one for suffering Tacoma owners).

Airbags

It’s estimated that about 50,000 airbags are stolen annually in the U.S. They’re valuable, fetching up to $200 on the black market, which makes them attractive to thieves. Thieves usually target dealerships or other places where cars are left unattended in accessible lots, but cars parked on the street are also hit.

How to stop thieves: The best ways to stop airbag theft is to lock your car when you park it and park it inside a secured area, like a garage. If that’s not possible, using a steering wheel lock using the “club” design can be effective, as it physically makes it difficult to access the driver’s-side airbag. This won’t stop thieves from grabbing the passenger-side airbag, of course, but cutting their profit in half might inspire them to move on.

License plates

Thieves steal license plates because it’s an easy way to cloak a vehicle. Sometimes they’re put on stolen vehicles, sometimes they’re used to evade tolls—but the fact that they’re generally easy to steal and very difficult to track makes them popular targets. In fact, the National Insurance Crime Bureau (NICB) notes that license plate theft has been rising rapidly—by 45% between 2019 and 2022 alone.

How to stop thieves: Swapping out the rusty, standard screws holding your plates on for some tamper-proof screws will do the job. These screws require a special fastener to attach or detach, and the kits come with screw covers to make it even more difficult to access.

Side mirrors

You might think your side mirrors are just mirrors, and who would bother stealing those? Ah, but modern-day side mirrors have a lot of technology inside them, especially on luxury vehicles, including blind spot sensors, defrosting systems, and cameras. Believe it or not, some mirrors can command upwards of $2,000 each!

How to stop thieves: If your side mirrors are motorized, turning them in when you park can make it more difficult for thieves to access them. You can also look for anti-theft mirror guards (like this one) for your make and model, which add a layer of security to the mirror to deter thieves.

Third row seats

If you have a truck or SUV with a removable third-row seat, you should know that it’s increasingly popular to steal them. The reason third-row seats are targeted is the universal reason anything gets stolen: These seats cost thousands of dollars to replace.

How to stop thieves: The best way to prevent theft of your truck’s third-row seat is to secure it with a bike lock. The specific lock that will work depends on how your third-row seat attaches and detaches, but all you need to do is secure it in place. It removes some of the convenience, but it’s a lot better than having your kids roll around in the cab while you drive.

Badges

People have been stealing car badges and hood ornaments for decades, often just to collect them or as petty acts of vandalism. But these days those emblems can actually contain a lot of expensive technology. Modern Volkswagens, for example, often tuck cruise control sensors into the hood emblem. That makes those badges a lot more valuable than you might think, and thieves are definitely aware of their added value.

How to stop thieves: It’s very difficult to stop someone from removing an emblem, badge, or ornament from your car. If you can access the back where it’s attached, you can try using wire or zipties to secure it more firmly—but if you can access it, so can thieves. Your best defense is to park the vehicle in a secure area whenever possible.

Secretary Rollins Announces Food Purchases for Communities in Need

(Washington, D.C., May 23, 2025) – U.S. Secretary of Agriculture Brooke Rollins today announced the U.S. Department of Agriculture’s (USDA) intent to purchase up to $67 million in fresh seafood, fruits, and vegetables from domestic producers to distribute to food banks and nutrition assistance programs across the country. These purchases are being made through Section 32 of the Agriculture Act of 1935 and will assist producers and communities in need.

The Powerbeats Pro 2 Are Perfect for the Gym and Are on Sale for Just $175 Right Now

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

If you’ve been eyeing workout-friendly earbuds that don’t flinch at sweat or bounce, the Beats Powerbeats Pro 2 are a great way to go. Originally released in February as a long-awaited follow-up to the first-gen Powerbeats Pro, these earbuds normally cost $249.99, but right now you can get a near-mint, refurbished pair on sale for $174.99 through StackSocial. They ship in “Grade A” condition, which means they may have barely-there scuffs and should look and work pretty much like new.

These are workout-first earbuds. The new nickel-titanium alloy hooks are flexible, lighter, and less clunky than before, keeping the buds in place without pressing too hard. They’re also IPX4 water-resistant, which is good enough for sweat or light rain, and you get a physical button to manage playback—a small detail, but one you’ll appreciate if your hands are wet or gloved. Wireless charging is now included, and you get around 8–10 hours of playback on a single charge, depending on whether ANC is on. That’s nearly double what most in-ears in this category offer (handy for fitness-first users who’d rather skip mid-run charging anxiety).

As for the caveats, the headline heart rate sensor sounds great on paper, but it underdelivers, especially if you’re hoping for smartwatch-level accuracy. And while the H2 chip (same as in AirPods Pro 2) helps with automatic pausing and connectivity, you’re still locked into Beats’ signature bass-heavy sound with no real way to tweak it through the app. If that’s your jam, you’ll likely love them, but if you prefer dialing in your own sound profile, you may feel boxed in. That said, for gym-goers or runners who want a solid, sweat-resistant set with long battery life and physical media controls, this refurbished deal hits the right balance between value and performance.

How the FDA Might Make It Harder to Get COVID Shots This Year

The U.S. government has not yet made its official recommendations for who should be able to get COVID booster shots this fall, but FDA officials published a policy position in the New England Journal of Medicine announcing that it intends to make some drastic policy changes. The changes could result in healthy people under age 65 losing access to COVID vaccines, according to vaccine experts who have spoken about the policies. Here’s what we know so far, and why the announced policy could be a problem. 

How COVID vaccines are currently approved

Scientists have changed the formulation of COVID vaccines a few times over the years, because the COVID virus itself tends to mutate. Vaccines are updated to better match the strains that are circulating, and this has happened roughly once a year—similar to how flu shots are updated each year. 

Instead of designing new vaccine trials from scratch for each small change in the COVID vaccine, manufacturers conduct studies to show that the immunity people get from the new vaccine is equivalent to what people got from the old vaccine. 

After approval from the FDA, the CDC then issues a recommendation for who should get the vaccine. Currently, everyone aged 6 months and up is recommended to get a COVID vaccine

What might be changing

The new policy, according to the NEJM article, would be to accept those immunobridging studies only to approve vaccines for people aged 65 and up, and people above the age of 6 months who have one of the high-risk conditions on a list maintained by the CDC. 

For healthy people under 65, the FDA’s policy wouldn’t approve new COVID vaccines unless they were tested against a placebo. (The type of placebo is phrased vaguely: “The control group could receive a saline placebo,” the authors write.) 

The FDA doesn’t have the authority to change the recommendations on who should get vaccines that are already approved (that’s the CDC’s purview), but it is in charge of approving vaccines and can approve them only for specific populations. 

Why placebo-controlled trials are an absolutely wild idea for COVID vaccines

Public health experts are, to put it mildly, not happy with this plan. That’s because we already have COVID vaccines that work. Doing a placebo-controlled trial would require withholding COVID vaccines from people in the control group; they would get saline instead of a functional vaccine. 

The normal way to do this type of trial (if you do one at all, rather than relying on immunobridging) is to compare the new vaccine or medication against one that is already considered effective. To use an extreme analogy, you wouldn’t test a new design of seatbelt by randomizing people to ride around without using any seatbelts at all. 

Vaccine scientist Peter Hotez told CNN that the FDA’s announced approach “essentially denies access to vaccines,” since such trials are not practical for companies to do. In a post on Bluesky, toxicologist Ryan Marino said that it amounts to “scientific misconduct.” Vaccine expert Paul Offit told NPR “I don’t think it’s ethical, given that we have a vaccine that works, given that we know that SARS-CoV2 [the COVID virus] continues to circulate and cause hospitalizations and death, and there’s no group that has no risk.”

More vaccine chaos may be coming

The new policy isn’t official yet, but it’s hard to imagine the FDA and CDC being allowed to approve and recommend vaccines the way it always has in the current political climate. Biologics director Vinay Prasad and FDA Commissioner Marty Makary, whose names appear on the FDA’s policy statement, have a history of arguing against COVID vaccine access for children. 

And both agencies are under the umbrella of HHS, the department of Health and Human Services, which is headed by Robert F. Kennedy, Jr—the same person whose anti-vaccine organization financed the movie Plandemic. If you don’t recall the details of that movie circulating in the early pandemic days, it implied both that COVID wasn’t real and that it was a bioweapon created by the government; the logic didn’t hold together but ultimately the point was that we should be suspicious of vaccines. (I have more on Plandemic here.) 

RFK, Jr has said a lot of bananas stuff about vaccines. He has compared childhood vaccines to the holocaust, claimed that Bill Gates put microchips in vaccines, and loudly questioned whether vaccines cause autism. How this man got put in charge of a health agency, I will never understand. 

Recent and future vaccine approvals may be at risk in this environment. Moderna had planned to submit a combined flu/COVID vaccine for approval; it has since withdrawn its application. (It’s not clear whether recent FDA policy announcements are directly related.) Novavax’s recent vaccine was approved recently, but only after a delay and only for older adults and for people with high-risk health conditions. Kennedy released a report today that questions the childhood vaccine schedule and implies that vaccines are part of the “stark reality of American children’s declining health.”